RunTests::check_if_equal for floats/doubles cannot distinguish between relative and absolute error

After fixing #183 at PR #198, we now do
```C++
    const double diff = fabs(b-a);
    if (diff <= tolerance)
            return true;

    const double largest = (std::max(fabs(b), fabs(a)));
    return ( diff <= tolerance*largest);
```
However, in many cases for floats, you might just want to do a relative error, where the 2 tolerances really could have very different values.
